SpaceX Internet Satellite Network: Starlink

[mongo]

A regular launch cadence for scheduled replacement, plus any special replacements, maybe hitching a ride, could make sense. Are there any technicalities on having to put replacement satellites into specific orbits that would make it difficult to shove a hitchhiker in there?

So out of my depth here, but I'll take a swing:
I think the normal method is to have the spares in a slightly different altitude. Then orbital prescession slow shift them though all the same inclination planes. Once it needs to take a sport it adjusts the normal altitude. With the Starlink, they might need to adjust the position of the other sat in the plane to get the gap to line up, but that can be a long duration low velocity maneuver.

Long term, they may just let the constellation grow and shrink as new ones launch and old ones are retired with that smae slow shifting of positions for uniformity.

 

[bxr140]

This statement ignores launching multiple sats at once, which it a huge cost save. Even Iridium was sub ten million (~6 million) a sat at a quantity of 10.

"Launch and maintain" implies a larger rack up of costs, not just the launch cost. If I had to provide a SWAGgy data point, I'd say the total per-unit recurring cost for Iridium is on the order of $20-30M per unit, maybe even more. That's launch cost, averaged unit cost, and operations cost. Fold in development and that number goes up even more.

Certainly from a feasibility perspective starlink needs to be at least an order of magnitude lower—probably a few million per unit, maybe even less. That’s still many many billions...

Here’s another quote from that article: “Each Starlink satellite will probably only last a few years, so SpaceX will need to launch new satellites to replace dead ones every few weeks.”

I’m not sure that is a well-founded assumption.

That's actually probably not too far off. One of the ways you drive down the cost of a satellite is to not use space rated parts (basically parts that are qualified to withstand the radiation environment), but that also decreases total lifetime. 5-7 years is the typical LEO lifetime for a low-ish cost satellite that still uses at least some high reliability parts. I’d be really surprised if design life of starlink is any more than 5 years. If they can get the per unit cost low enough it might even be 3...then hope for 4 or 5.

That maths out to a lot of failures annually; certainly the plan would be to launch a ton of units at once (instead of onsey-twosie as units fail) and maintain a fleet of on-orbit spares. The volume is going to be high enough that the whole attrition-replacement plan should be pretty predictable.

 

[ggr]

Article includes some info about Starlink:
SpaceX reveals new details about its biggest, most secretive projects

 

[ecarfan]

Thanks. Yes, some info, but despite the clickbait title no “new details”. And some significant misrepresentations by the writer.

Article includes some info about Starlink:
SpaceX reveals new details about its biggest, most secretive projects

 

[bxr140]

Article includes some info about Starlink:
SpaceX reveals new details about its biggest, most secretive projects

Yikes!

Besides being a pretty solid article, unless I missed previous news it’s an absolute bombshell to drop the altitude by ~half.

 

[ggr]

Yikes!

Besides being a pretty solid article, unless I missed previous news it’s an absolute bombshell to drop the altitude by ~half.

I believe much of the constellation is getting closer, but they're going to put some even further out for the long-distance backhaul.

 

[BioSehnsucht]

I believe much of the constellation is getting closer, but they're going to put some even further out for the long-distance backhaul.

Do you mean still put some out at the original LEO altitude (~1150km or something like that, IIRC these new docs show ~half at ~550km and ~half at ~1150km or something along those lines) ? Or you expect them to put some out even further than that (if so, why? does not seem to provide any advantages ... ? )

 

[ggr]

Do you mean still put some out at the original LEO altitude (~1150km or something like that, IIRC these new docs show ~half at ~550km and ~half at ~1150km or something along those lines) ? Or you expect them to put some out even further than that (if so, why? does not seem to provide any advantages ... ? )

I can't remember or find a reference, but basically if you put some out higher, they have line-of-sight (Line-of-Laser LoL!) further, so less hops for intercontinental traffic.

 

[Grendal]

Our Ludicrous Future did a segment on Starlink. Mini-BFR at 12:00. Starlink at 49:30.

 

[BioSehnsucht]

I can't remember or find a reference, but basically if you put some out higher, they have line-of-sight (Line-of-Laser LoL!) further, so less hops for intercontinental traffic.

Fewer hops, but further distance. So depending on the throughput capabilities of the satellites, it might be slower to do so.

I've seen many assume that routing delays will be "bad" (for the number of hops), because historically they have been, but I wouldn't assume that. I'd expect the routing throughput to be designed for 100% saturation of both up an down links to Earth as well as laser links to other satellites - though this is probably at some assumed average packet size, vs scaling for SYN floods. I expect that they won't natively operate using BGP or other typical IP based routing methods but something optimized to their use case, as the best destination (and thus best route) to reach will be changing constantly (overall a route around the world is likely changing every few seconds if not more often, and the best satellite to route to before downlinking to a ground station / user terminal will change every few minutes at least). By designing something purpose made for it, I expect it to perform better than many existing routers on Earth do (as many of them are over saturated because the bandwidth to/from them scales faster than they are replaced with more capable routers, which causes buffering and slowdown).

The exception of course (assuming the routing can keep up with throughput) being that at some point you hit maximum throughput, or maximum throughput for an area of the globe, and can't fit more satellites in the same altitude so must add more altitudes to gain more bandwidth.

 

[mongo]

Fewer hops, but further distance. So depending on the throughput capabilities of the satellites, it might be slower to do so.

I've seen many assume that routing delays will be "bad" (for the number of hops), because historically they have been, but I wouldn't assume that. I'd expect the routing throughput to be designed for 100% saturation of both up an down links to Earth as well as laser links to other satellites - though this is probably at some assumed average packet size, vs scaling for SYN floods. I expect that they won't natively operate using BGP or other typical IP based routing methods but something optimized to their use case, as the best destination (and thus best route) to reach will be changing constantly (overall a route around the world is likely changing every few seconds if not more often, and the best satellite to route to before downlinking to a ground station / user terminal will change every few minutes at least). By designing something purpose made for it, I expect it to perform better than many existing routers on Earth do (as many of them are over saturated because the bandwidth to/from them scales faster than they are replaced with more capable routers, which causes buffering and slowdown).

The exception of course (assuming the routing can keep up with throughput) being that at some point you hit maximum throughput, or maximum throughput for an area of the globe, and can't fit more satellites in the same altitude so must add more altitudes to gain more bandwidth.

Indeed. Since they control all nodes, the routing algorithm is a complete knowledge system, not an interconnected mish mash of providers and links.

 

[bxr140]

I believe much of the constellation is getting closer, but they're going to put some even further out for the long-distance backhaul.

Yeah, for sure multiple altitudes makes sense...that’s been the plan for a while...I think? And adjusting the altitude over the course of development is also expected.

Most curiously, the new altitude is more or less where the pathfinders are right now.

It wouldn’t surprise me if insertion and deorbiting play a part in this change. On the front end, not having to raise the orbit after separation saves time and, especially over a truckload or satellites, launch mass, because each unit needs incrementally less fuel.

On the back end, especially given the typical and likely shorter than typical life of a starlink, allowing the satellite to naturally re-enter is going to REALLY simplify logistics, and again save fuel.

 

[Ed Hart]

Things DO change, Mongo! Thanks.
There will be a day when an IPO makes sense. At the moment, there is enough private money to keep things moving forward.
Great company.

 

[mongo]

Things DO change, Mongo! Thanks.
There will be a day when an IPO makes sense. At the moment, there is enough private money to keep things moving forward.
Great company.

Yah, down the road. SpaceX is doing a 750 million raise of private funding. That will help fund inital satellite builds. The BFR will lower lauch costs incredibly. So SpaceX will be able to get some sat coverage up. Once they have a minimal functional constellation, they can get subscribers. That will help fund more sats, and the cycle becomes self sustaining.
An IPO pushes SpaceX into a for profit company instead of a for mission one. So they would need to increase prices above the neccessary level to achieve the addition revenue back to the shareholders or else reduce the rate of investment/ expansion.

Until they approach saturation. At that point they could IPO based on their established revenue streams and use that to fund more BFRs sooner (assuming production capacity) . Like a loan you never repay the principal on.

 

[Model 3]

An IPO pushes SpaceX into a for profit company instead of a for mission one. So they would need to increase prices above the neccessary level to achieve the addition revenue back to the shareholders or else reduce the rate of investment/ expansion.

Until they approach saturation. At that point they could IPO based on their established revenue streams and use that to fund more BFRs sooner (assuming production capacity) . Like a loan you never repay the principal on.

But the "Starlink" part is about generating profit back to SpaceX so they can get on with their mission. So they could, if they needed the founding, put Starlink into a company of it's own, and then do a IPO on that.

 

[mongo]

But the "Starlink" part is about generating profit back to SpaceX so they can get on with their mission. So they could, if they needed the founding, put Starlink into a company of it's own, and then do a IPO on that.

Sure, but if Starlink is seperate from SpaceX, does that impact the ability of Starlink to hire SpaceX for launches and what rate they are charged? Not sure if that is a concern, just thinking of the Solar City/ Tesla battery issues.

Other funding option is pre-paid service plans. 5 years of coverage at a discount. Once the infrastructure is in place, the data packets are free. (All fixed costs, other than non Starlink backbone connections)

 

[electracity]

I can't remember or find a reference, but basically if you put some out higher, they have line-of-sight (Line-of-Laser LoL!) further, so less hops for intercontinental traffic.

Higher orbit for backhaul would eliminate the low latency advantage of the low orbit satellites. It seems to me that the low satellites must normally make a direct internet connection to a ground station.

There are perhaps multiple service offerings here. Low latency/high bandwidth where there are appropriate ground stations. But higher latency "Sat phone" type performance everywhere else.

They can also add satellites to the orbit for high demand areas as the market develops. The 7000 satellite number is probably not for a uniform constellation at all.

I would love to know the potential bandwidth of one of these satellites. Anyone know what throughput a single high end cell tower can handle?

 

[Model 3]

Sure, but if Starlink is seperate from SpaceX, does that impact the ability of Starlink to hire SpaceX for launches and what rate they are charged?

First I must say that I reelly do not know. But as long at SpaceX and/or Elon have the controlling majority in "Starlink company" I do not see that as a problem.

... but on the other hand: If Starlink find that they can get more profit/get the constellation going earlier by spreading out to different launch operators, would that really matter? The point is to get it going and generate profit to it's owners (SpaceX).

Other funding option is pre-paid service plans. ...

Yes, this is not the only way, and I don't even know if I would recommend it. I just pointed out that Starlink could be public while still have SpaceX private.

 

[mongo]

Higher orbit for backhaul would eliminate the low latency advantage of the low orbit satellites. It seems to me that the low satellites must normally make a direct internet connection to a ground station.

There are perhaps multiple service offerings here. Low latency/high bandwidth where there are appropriate ground stations. But higher latency "Sat phone" type performance everywhere else.

They can also add satellites to the orbit for high demand areas as the market develops. The 7000 satellite number is probably not for a uniform constellation at all.

I would love to know the potential bandwidth of one of these satellites. Anyone know what throughput a single high end cell tower can handle?

Higher orbit does not need to be a huge distance increase, the higher orbit sat can be on the path to the destination. Could go one low to one high to one low for instance. Instead of multiple low hops to get around the Earth.

The sats are not geosynchronous, so you can't only add coverage to one area by boosting numbers. The low altitude will reduce spot size allowing a higher number of sats to service a given area.

 

[ggr]

Higher orbit for backhaul would eliminate the low latency advantage of the low orbit satellites. It seems to me that the low satellites must normally make a direct internet connection to a ground station.

There are perhaps multiple service offerings here. Low latency/high bandwidth where there are appropriate ground stations. But higher latency "Sat phone" type performance everywhere else.

They can also add satellites to the orbit for high demand areas as the market develops. The 7000 satellite number is probably not for a uniform constellation at all.

I would love to know the potential bandwidth of one of these satellites. Anyone know what throughput a single high end cell tower can handle?

You're not understanding the order-of-magnitudes here. The low satellites are around 500km, while the higher ones are around 1400km (approximating, from memory). Berlin to Chicago, an approximation for average Europe-US distance, is about 7000km. So the high satellites are significant detours, adding about 40% to the distance, just about compensating for the speed-of-light savings. But Clarke Orbit satellites are 40,000km away. A single round trip to next door, via geosynchronous orbit, is more than 10 times as far as Europe-US.